Cross-Country Evidence on Public Sector Retrenchmentppiaf.org/sites/ppiaf.org/files/documents/toolkits/...John Haltiwanger and Manisha Singh are with the Department of Economics at

23

THE WORLD BANK ECONOMIC REVIEW, VOL. 13 , NO. 1 : 23–66

© 1999 The International Bank for Reconstruction and Development / THE WORLD BANK

John Haltiwanger and Manisha Singh are with the Department of Economics at the University ofMaryland. This is the revised version of a paper presented at the World Bank conference on PublicSector Retrenchment and Efficient Compensation Schemes held in November 1996. The authors aregrateful to many officials at the World Bank for providing data and their time for extensive interviewson specific public sector retrenchment programs. They thank participants at an informal seminar held atthe World Bank in January 1996, participants at the conference in November, Martín Rama, PeterFallon, and the anonymous referees for helpful comments. Catherine Buffington and Andrew Figuraprovided excellent research assistance.

Cross-Country Evidenceon Public Sector Retrenchment

John Haltiwanger and Manisha Singh

This article reports the results from a survey of public sector employment retrench-ment episodes across a wide variety of developing and transition economies. The in-formation collected and analyzed is primarily from internal World Bank documentsand in-depth interviews with World Bank staff having operational information aboutexperiences in specific countries. Using the information collected on 41 retrenchmentprograms across 37 countries, the article analyzes the relationships between the fac-tors leading to retrenchment, the scope and nature of retrenchment, and the methodsused to accomplish the retrenchment. The discussion of methods includes an analysisof the mix of involuntary and voluntary employment reduction programs, the com-pensation schemes offered, and the extent of targeting of specific types of workers.Although relevant quantitative information is limited, the article also attempts to evalu-ate the outcome of the programs on several dimensions. The most striking findingsrelate to analysis of the factors leading a significant fraction of programs to rehireworkers separated from the public sector (thereby defeating the programs’ objective).In addition, the article relates program characteristics to calculated summary financialpayback indicators and to the nature of the labor market adjustment.

The retrenchment of public sector employment is an increasingly pervasive phe-nomenon. Many industrial, developing, and transition economies have recentlyfaced the issue of downsizing their public sector employment. The reasons un-derlying the downsizing vary considerably across countries. For some, it is ageneral move toward a more market economy; for others, it is a reduction in therole of the military or an attempt to reduce a bloated bureaucracy; for others, itis sparked by a fiscal crisis necessitating a severe cutback in government spend-ing; and, finally, for some, it is a combination of some or all of these.

Given the pervasiveness of the phenomenon, there is no shortage of studies ofindividual countries or episodes. Although country-specific studies are quite use-

24 THE WORLD BANK ECONOMIC REVIEW, VOL. 13, NO. 1

ful, it is difficult to draw common lessons from them, and there is relatively littleanalysis comparing experiences across countries. Important exceptions are Svejnarand Terrell (1991), which provide a useful comparison across six countries forretrenchment in the transport sector, and Lindauer and Nunberg (1994), whichprovide a book-length discussion of civil service reform. Our study has benefitedsubstantially from the insights in both of these publications. In this article, wecompile and analyze information on the recent retrenchment experiences acrossa large number of countries and retrenchment episodes. We take a broad view ofpublic sector employment and associated retrenchment. Public sector employ-ment includes civil service workers, military personnel, and employees of publicsector enterprises.

Comparing and contrasting the experiences across countries are formidabletasks. Beyond the usual problems of cross-country comparisons, no central da-tabase has compiled the requisite information that permits such analysis. Ac-cordingly, we compiled and collected our own information from myriad sources,but primarily from internal World Bank documents and interviews of WorldBank staff having operational information about retrenchment experiences inindividual countries. (Section II provides details regarding the collection of in-formation; a complete listing of individual country summaries is available fromthe authors upon request.)

Our analysis is based on 37 countries and 41 retrenchment programs. Mostof the programs for which we could collect detailed information commenced inthe early 1990s, and many of the programs are ongoing (which makes analysisof the ultimate success of the programs difficult). The gross number of workersseparated in all the programs considered exceeds 5 million, with a somewhatsmaller net reduction in employment. The discrepancy reflects the fact that someprograms exhibited significant rehiring of separated workers or hiring of newworkers by the public sector. In the analysis that follows, we look closely at thecharacteristics of programs that involved significant rehiring and new hires. Al-though new hires may be part of a coherent plan to restructure the public sectorwork force, rehiring is clearly not an indicator of success. The converse is nottrue; that is, the absence of rehiring is not synonymous with success. Also, arestructuring could involve separating employees en masse and subsequentlyrehiring those with necessary specific skills or rehiring workers on temporaryterms or both. No program in our sample intended such purposive rehiring.

Evaluating the costs and benefits of individual programs in a more compre-hensive fashion is inherently difficult. In principle, the requisite information in-cludes the compensation packages offered, the relative productivity and wagesof the displaced workers in the public and private sectors, and the adjustmentcosts borne directly by the affected workers and the entire economy. Measure-ment of much of this—particularly the nature of the adjustment costs—is wellbeyond the scope of available information. Nevertheless, we can document con-siderable details about the magnitude and forms of compensation used in re-trenchment programs and the wage bill savings, and we can characterize some

Haltiwanger and Singh 25

of the other aspects that are relevant for evaluation. To summarize the availableinformation on costs and benefits, we calculate a simple financial break-evenindicator measured as the number of years required for the present value offinancial costs to equal the present value of financial gains. We also examinerefinements of this indicator that attempt to incorporate the productivity gainsgenerated from the retrenchment. In turn, we relate these summary financialindicators to other program characteristics.

Section I provides a brief presentation of the underlying conceptual frame-work for characterizing the private and social costs and benefits of a publicsector retrenchment program. The analysis is deliberately simple and borrowsheavily from the existing literature. Our primary objective is to provide guid-ance about the type of information that is required to compare and analyzealternative programs and to discuss the interaction between the conceptual andmeasurement issues that must be confronted. Section II outlines the methodol-ogy used to collect information for the individual countries and to compile theinformation in a systematic fashion. Section III presents the analysis of the infor-mation collected as well as some detailed discussion of individual countries toillustrate the patterns that emerge. Section IV offers concluding remarks.

I. CONCEPTUAL FRAMEWORK AND MEASUREMENT ISSUES

A simple conceptual framework helps us to organize the information we havecollected and provides some guidance on the type of factors relevant for com-paring and analyzing retrenchment programs across countries. Although the dis-cussion in this section borrows heavily from the existing literature, our charac-terization of the relevant issues emphasizes some points that are neglected in theliterature. For example, many of the conceptual issues discussed in this sectionare discussed more formally and with greater attention to the range of relevantissues in Diwan (1993a, 1993b) and Rama (1997). Because our ultimate objec-tive is to quantify the relevant measures using data across countries, we focus onthe interaction between conceptual and measurement issues.

The analysis considers six relevant private and social incentives for the re-trenchment decision regarding the marginal worker. Although we do not char-acterize all relevant sources of heterogeneity explicitly, differences across work-ers in the following present values may reflect differences in ability, experience,skills, horizons, discount rates, and mobility costs. Pprivate is the present discountedvalue of the worker’s productivity in the private sector. Ppublic is the presentdiscounted value of the worker’s productivity in the public sector. Cindividual is thepresent discounted value of adjustment costs borne by the individual in relocat-ing from the public to the private sector (for example, job search costs, reloca-tion costs, and time spent unemployed). Csocial is the present discounted value ofadjustment costs borne by society for the individual to relocate from the publicto the private sector (that is, adjustment costs borne by the individual plus spillovereffects such as congestion effects). Wpublic is the present discounted value of the


earnings of the worker in the public sector. Wprivate is the present discountedvalue of the earnings of the worker in the private sector.

The discussion here and the subsequent analysis focus primarily on the reallo-cation of workers from the public to the private sector taking into account therelevant productivity, wages, and costs of labor market adjustment. Public sec-tor downsizing also likely involves the reallocation of capital. Even though ourfocus is on the reallocation of workers, the implications of capital reallocationand its interaction with worker reallocation deserve further attention. Also, ex-plicit modeling of the distinction between the present discounted value of earn-ings and adjustment costs would require incorporating the fact that transitionsfrom a public sector job to a private sector job may involve several transitionsand, accordingly, several spells of unemployment. As emphasized by Hall (1995),a potentially important explanation of persistence in unemployment rate dy-namics is that separations tend to beget further separations.

The principle of social optimality considered here is that all resources shouldbe allocated to their highest-valued use (net of relocation costs given the existingallocation of resources). Thus it is socially optimal to relocate the marginal workerto the private sector if:

(1) Pprivate – Csocial > Ppublic.

An individual worker will choose to stay in a public sector job as long as:

(2) Wpublic > Wprivate – Cindividual.

If workers are paid more than their marginal product in the public sector(that is, Wpublic > Ppublic), then it is easy to imagine that inequalities 1 and 2 canboth hold simultaneously. That is, it is in society’s interest for the marginalworker to relocate, but it is not privately optimal.

It may be that individual workers and the government discount the futuredifferently because of differential access to capital markets. Such differentialaccess to capital markets is, in principle, consistent with the specification andassociated discussion considered here, but explicit consideration of the role ofcapital markets deserves more attention. For example, unemployment benefitsand other forms of worker safety net assistance are often justified as a form ofsocial insurance in the face of imperfections in the capital market. Because en-hancements of the worker safety net are often part of a retrenchment program,explicit consideration of the role of differential capital market access is quiterelevant.

Several additional factors are potentially important in evaluating the optimalityof a retrenchment program. One factor is the expenditures incurred to pay thewages of public workers or, alternatively, the compensation packages offered toinduce workers to relocate voluntarily. If taxes are not distortionary and there isno distortionary rent-seeking behavior, these expenditures should be viewed astransfers and thus should not affect efficiency. Under these strong assump-tions, these terms should not appear in inequality 1. However, in the presence of


distortionary taxes (including the inflation tax) and distortions from rent-seeking behavior, such transfers yield efficiency losses. In addition, an excessivewage bill or fiscal crisis may imply that retaining redundant workers may ad-versely affect other government services.

To incorporate these effects in a modified version of inequality 1, supposethat there is a distortions-based loss function L(.) that is an increasing functionof the transfers to workers. Taking into account these distortionary losses ininequality 1, retrenchment is optimal if:

(3) Pprivate – Csocial – L(COMP) > Ppublic – L(Wpublic)

where COMP reflects the present discounted value of the compensation or otherassistance programs that accompany the retrenchment. A related issue is whetherthe nature of the distortions from the transfers depends on the timing of theprogram as well as on the net present value of the transfers. For example, rentsin the form of artificially high public sector wages may have dynamic distortionaryeffects beyond those associated with a one-time transfer.

Policymakers face the problem (shared in our analysis) of measuring the com-ponents in inequalities 1, 2, and 3. Many of the components are difficult tomeasure or difficult to observe in the presence of worker heterogeneity (outsideopportunities and individual productivity). Among the most difficult to assessare the adjustment costs. Factors influencing adjustment costs include the degreeof labor market flexibility (barriers to adjustment in terms of hiring and firingcosts), the worker safety net, informal compared with formal sector develop-ment, and the method, scope, and speed of retrenchment.

Social adjustment costs will be greater than private adjustment costs to theextent that there are spillover or externality effects of worker displacement. Thespillover and externality effects include the impact of changes in the demand forfinal goods due to reduced consumption expenditures by affected workers, con-gestion externalities among job searchers, and social disruption (such as na-tional strikes). For further discussion of the congestion externalities among jobseekers, see, for example, Blanchard and Diamond (1989, 1990). For furtherdiscussion of the idea that employment and earnings losses by workers in onesector may generate spillover effects in other sectors by changing the demand forfinal goods, see, for example, Cooper and Haltiwanger (1996). The method,scope, and speed of retrenchment potentially influence these spillover effects. Inaddition, the concentration of the retrenchment in a local community may implyimportant local spillover effects even if there are no economywide spillovereffects.

The nature and magnitude of these adjustment costs are very difficult to mea-sure, and we have relatively little quantifiable information on adjustment coststhat we can use in our comparison across countries. The most direct evidenceavailable is from recent studies of privatization in transition economies for whichthe massive restructuring implies that adjustment costs take center stage (see, forexample, the studies in Commander and Coricelli 1994). Even in these studies,


information on the nature of the adjustment process is often quite limited. Ingeneral, and particularly in the transition economies, characterizing these ad-justment costs is at the heart of the debate regarding gradualism and the bigpush in efforts to reduce the role of the public sector in the economy.

Another difficult measurement and conceptual problem is evaluating the pre-and post-retrenchment productivity of workers in the public sector. Althoughthere may be widespread agreement that some public sector workers are redun-dant, quantifying this is extremely difficult. The nature of the measurement prob-lems in evaluating pre- and post-retrenchment productivity will vary dependingon the type of public sector employment (civil service, military, or public sectorenterprises producing goods and services). Further, some of the need for re-trenchment may reflect more of a need for restructuring than for simplydownsizing. For example, low morale and low productivity of public sectoremployees may reflect poor organizational structure or wage compression. Thusa program may involve changes in the organizational structure and the wagestructure that appear to be financially expensive but are well motivated by suchfactors. Simple financial indicators that do not adequately incorporate such fac-tors should not be used to evaluate the need for or success of a program. Interms of inequality 3, this discussion implies that the public sector productivityterm needs to be interpreted broadly. That is, retrenching the marginal workerwill yield the direct loss of the worker’s output (if any) but may also yield pro-ductivity gains from the reorganization that accompanies downsizing.

Beyond the problems of measurement and assessment, it is often politicallynecessary to implement the retrenchment scheme through a voluntary program.This necessitates the provision of some incentive payment in the form of sever-ance pay or pensions (denoted INCENTIVE below) such that the following in-equality would hold for an individual worker:

(4) INCENTIVE > Wpublic – (Wprivate – Cindividual).

In principle, with complete information about worker heterogeneity, it is pos-sible to design an optimal incentive scheme for each worker (see Diwan 1993a,1993b and Rama 1997 for a formal analysis). One well-recognized problemthat immediately emerges is that if a simple, common incentive package is of-fered to all workers, heterogeneity across workers implies that only those withthe lowest rent will depart. It is important to emphasize that some aspects of theselection process may be unfavorable, while others may be advantageous. Forexample, in environments with wage compression, a common package offeredto a wide class of workers implies that the most skilled and capable workers willleave. Individuals with better outside opportunities or low adjustment costs aremore likely to leave. Other things being equal, this voluntary aspect of selectionwill be favorable.

The implication of this discussion is that an optimal incentive scheme must beindividually tailored to reflect worker heterogeneity in rents and adjustment costsbut that this is difficult to implement in practice in the face of imperfect infor-


mation. This tension is at the center of the debate about the pros and cons ofvoluntary and involuntary retrenchment programs. As noted, voluntary pro-grams have the advantage that they yield some favorable voluntary selection ofworkers with good outside opportunities or low adjustment costs without re-quiring the policymaker to have complete information. However, workers withgood outside opportunities are likely the most productive public sector workers,so that the voluntary selection may be adverse. The problem with adverse selec-tion is apt to be especially severe in environments with wage compression. Levyand McLean (1996) discuss optimal schemes using different types of severancecontracts or auctions.

Involuntary programs potentially permit specific targeting of groups of work-ers on observable characteristics but do not take advantage of favorable volun-tary selection on unobservable characteristics. Further, involuntary schemes mayalso yield high adjustment costs. For example, mass involuntary layoffs mayinvolve substantial private and social adjustment costs, and in this case the na-ture of the safety net takes on particular importance. Finally, the political will(ability) to undertake and sustain schemes with an involuntary component maybe lacking.

Putting these pieces together suggests that our comparison and analysis ofalternative programs should consider the following factors. First, we must con-sider the factors leading to retrenchment (fiscal crisis, overstaffing, low moraleas a result of wage compression) because they may provide insights about therelationship between wages and productivity in the private and public sectors.Second, several factors influence the adjustment costs, including the scope andspeed of retrenchment, the mechanism used (involuntary or voluntary), the useof targeting (for example, based on skill or age), and the nature of labor marketflexibility and the safety net. Many of these same factors are important (espe-cially the method used and the nature of targeting) in terms of the need forindividual tailoring of plans given worker heterogeneity. Finally, the magnitudesof the financial costs (Wpublic and INCENTIVE) are relevant for characterizingthe magnitude of the transfers.

II. DATA COLLECTION

The survey of public sector retrenchment programs across developing andtransition countries was carried out on two dimensions. First, we collected inter-nal World Bank documents relating to macroeconomic and public sector adjust-ment in these countries. We drew up a preliminary portrait of the issues relatingto each country using various World Bank documents.1 We used a variety ofexternal sources to supplement these documents. Second, we interviewed World

1. The World Bank documents included Staff Appraisal Reports, Memoranda and President’sRecommendations, President’s Reports, supervision memoranda, Project Completion Reports,Implementation Completion Reports, Project Performance Audit Reports, Operations Evaluation Studies,Country Economic Memoranda, and sector and other economic reports.


Bank officials associated with the retrenchment programs to obtain more directinformation and assessment. The interviews, conducted over more than eightweeks during the summer of 1995, typically yielded further acquisition of rel-evant documents. The countries, retrenchment programs, and associated offi-cials were selected using the World Bank’s electronic management informationsystem and the adjustment lending database called ALCID. Appendix table A-1lists the surveyed countries and programs. The projects and loans listed in theWorld Bank’s management information system go back to the mid-1980s. Theselection of programs was mostly restricted to this list. In general, the movementof World Bank staff away from early programs reduced the usefulness of suchinterviews.

The interview transcripts, documents, reports, and articles relating to eachprogram were synthesized and summarized using a uniform outline. The infor-mation presented relates to four broad aspects characterizing retrenchment, thenature of labor turnover and institutions, cost-benefit analyses in financial andeconomic terms, and monitoring and evaluation of the program. A completelisting of the individual country summaries is available from the authors uponrequest.2

The information collected and assimilated ranges from quantitative informa-tion about scope, speed, and financial costs and benefits to qualitative informa-tion about the factors precipitating the retrenchment, as well as characteristicsof the program such as the methods used and the productivity gains observed inthe public sector. The quality and completeness of the information collectedvary substantially across programs. This blend of quantitative and qualitativeinformation yields an analysis that is primarily descriptive—an attempt at sum-marizing the basic facts and drawing out observable patterns. Not surprisingly,a host of institutional and idiosyncratic factors appear to be important. Theseinstitutional and idiosyncratic factors serve as an additional reminder to inter-pret the analysis of basic patterns with caution.

III. SURVEY AND ANALYSIS OF CROSS-COUNTRY EVIDENCE

We begin the analysis by characterizing some basic facts about the programsfor which we gathered information. Tables 1 and 2 present summary statisticsabout the programs surveyed (summary information on each program is pro-vided in the cross-country appendix tables). The analysis is based on 41 pro-grams in 37 countries. The World Bank has not historically provided direct as-sistance for public sector employment retrenchment programs. (As of February1997, it allowed lending for severance payments provided they entail no organi-zation closure.) However, under the umbrella of a comprehensive assistancepackage, an agreement with a country often stipulates that the country use do-mestic counterpart funds for specific purposes. Under such umbrella agreements,

2. The detailed country summaries total more than 200 pages, including extensive references todocuments, papers, and other sources beyond those cited here.


public sector retrenchment may be part of the overall package. In our survey,about 65 percent of the programs had this umbrella connection to World Bankfinancing.

Table 1 shows that total separations are greater than the total reduction inemployment, reflecting rehiring and new hires. Most programs did not experi-ence rehiring of the same workers who had departed, but a nontrivial fraction(20 percent) experienced significant rehiring. New hires were also relatively rare(about 13 percent of programs). Some programs represented only good inten-tions, with no workers actually separated. One of the programs included in oursurvey is Hungary’s massive public sector retrenchment program. This programis a clear outlier on a number of dimensions, but we felt it was important toinclude large programs among the transition economies as points of contrast.

In table 1, the method of employment reduction is divided into three basiccategories: involuntary-hard refers to layoffs; involuntary-soft refers to employ-ment reductions generated by strict enforcement of rules such as mandatoryretirement and the removal of ghost workers (workers on the payroll who donot exist, although someone is collecting the paycheck); and voluntary refers toprograms in which employment reductions were achieved through workerswho quit voluntarily (for example, early retirements). Table 1 shows thatinvoluntary-hard reductions dominate total reductions in employment. How-ever, this is primarily driven by the massive involuntary reductions in EasternEurope. By contrast, the use of voluntary and involuntary-soft measures is theprevailing pattern in Latin America, Asia, and Africa.

Table 2 indicates that the total financial costs of the retrenchment programswere large (exceeding $12 billion), with the average program having a total costof $400 million. The total cost is driven in part by enormous increases in pen-sion and safety net expenditures in Poland. From all accounts, the increases forPoland reported in appendix table A-2 are real, but they skew totals and aver-ages somewhat. However, much of the subsequent analysis is not sensitive tooutliers on this dimension. That is, much of the analysis is based on the percent-age of programs that exhibit various characteristics.

The financial costs of the programs took a variety of forms, including sever-ance payments, higher pension liabilities, and enhancements to the worker safetynet. The worker safety net is a shorthand term for the various worker assistanceprograms, including unemployment benefits, job search assistance, training, andrelocation assistance designed to aid in the process of worker reallocation. As isclear from the minimum and the maximum in table 2, the scope and mix ofpackages varied considerably across countries. As the patterns across continentsshow, the transition economies in Eastern Europe relied heavily on the workersafety net. Countries in the Southern Hemisphere and Asia relied much moreheavily on direct compensation to workers, such as severance pay and enhancedpensions. The latter pattern is particularly striking for severance pay, which was92 percent of total costs in Asia, 74 percent in Latin America, and 51 percent inAfrica.

32

Tab

le 1

. Sc

ope

of R

etre

nchm

ent:

Num

ber

of W

orke

rs, 1

990s

By

inst

rum

ent

Em

ploy

men

tIn

volu

ntar

y-In

volu

ntar

y-R

egio

nre

duct

ion

hard

soft

Vol

unta

rySe

para

tion

sR

ehir

esN

ew h

ires

Tot

alA

ll ca

ses

4,71

0,56

61,

178,

394

752,

809

577,

188

5,03

3,87

528

2,30

741

,002

Afr

ica

892,

062

104,

035

140,

896

97,8

7890

9,70

515

,900

1,74

3A

sia

233,

311

011

,000

172,

711

255,

311

3,00

019

,000

Eur

ope

2,85

3,55

988

3,25

913

2,00

010

5,00

02,

853,

559

00

Lat

in A

mer

ica

731,

634

191,

100

468,

913

201,

599

1,01

5,30

026

3,40

720

,259

Med

ian

All

case

s32

,900

01,

800

4,80

037

,500

00

Afr

ica

10,0

611,

922

6,86

13,

696

10,0

610

0A

sia

21,9

250

014

,373

23,4

250

0E

urop

e17

2,95

942

,000

017

,500

172,

959

00

Lat

in A

mer

ica

26,0

000

04,

599

56,4

090

0

Ave

rage

All

case

s11

7,76

439

,280

27,8

8218

,037

125,

847

7,05

81,

025

Afr

ica

59,4

7110

,404

14,0

907,

529

60,6

471,

060

116

Asi

a29

,164

02,

200

28,7

8531

,914

375

2,37

5E

urop

e40

7,65

117

6,65

233

,000

26,2

5040

7,65

10

0L

atin

Am

eric

a73

,163

19,1

1058

,614

22,4

0010

1,53

026

,341

2,02

6

Min

imum

All

case

s0

00

024

70

0A

fric

a24

70

00

247

00

Asi

a6,

500

00

06,

500

00

Eur

ope

35,0

000

00

35,0

000

0L

atin

Am

eric

a0

00

02,

034

00

Max

imum

All

case

s1,

661,

000

547,

300

424,

095

112,

000

1,66

1,00

016

3,05

919

,000

Afr

ica

541,

200

57,0

0075

,000

59,8

1054

7,20

07,

500

1,74

3A

sia

69,4

660

7,00

069

,466

69,4

663,

000

19,0

00E

urop

e1,

661,

000

547,

300

132,

000

70,0

001,

661,

000

00

Lat

in A

mer

ica

405,

995

100,

000

424,

095

112,

000

424,

095

163,

059

18,1

00

Not

e: V

alue

s ar

e ba

sed

on 4

1 re

tren

chm

ent

prog

ram

s in

37

coun

trie

s. I

nfor

mat

ion

abou

t w

orke

r se

para

tion

s by

inst

rum

ent

is in

com

plet

e in

sev

eral

cas

es.

Sour

ce: A

utho

rs’ c

alcu

lati

ons

base

d on

indi

vidu

al c

ount

ry s

umm

arie

s (s

ee a

ppen

dix

tabl

es f

or c

ross

-cou

ntry

tab

ulat

ions

).

33

Tab

le 2

. Sc

ope

of R

etre

nchm

ent:

Fin

anci

al C

osts

, 199

0sFi

nanc

ial c

osts

(m

illio

ns o

f do

llars

)Fi

nanc

ial b

enef

its

(mill

ions

of

dolla

rs)

Seve

ranc

eE

nhan

ced

Safe

tyC

ost

per

Ann

ual w

age

Reg

ion

Tot

alpa

ymen

tspe

nsio

nne

tw

orke

rT

otal

bill

savi

ngs

Tot

alA

ll ca

ses

12,0

742,

708

6,09

83,

268

n.a.

4,20

62,

834

Afr

ica

500

255

2022

5n.

a.67

396

Asi

a1,

400

1,28

171

48n.

a.17

9–1

53E

urop

e8,

583

05,

588

2,99

5n.

a.1,

846

1,44

6L

atin

Am

eric

a1,

591

1,17

241

90

n.a.

1,50

81,

445

Med

ian

All

case

s42

130

01,

085

1710

Afr

ica

2013

00

1,47

610

8A

sia

5025

024

1,04

09

9E

urop

e45

40

2585

961

692

372

3L

atin

Am

eric

a28

011

20

04,

735

222

222

Ave

rage

All

case

s40

287

244

192

4,14

216

812

3A

fric

a38

202

323,

630

6111

Asi

a20

018

314

243,

651

26–2

2E

urop

e2,

146

01,

397

998

3,81

792

372

3L

atin

Am

eric

a26

516

770

05,

695

302

289

Min

imum

All

case

s0

00

00

–157

–157

Afr

ica

20

00

320

00

Asi

a0

00

00

–157

–157

Eur

ope

60

06

2429

829

8L

atin

Am

eric

a2

00

00

–15

–15

Max

imum

All

case

s7,

669

1,14

05,

538

2,13

017

,108

1,54

81,

148

Afr

ica

200

8020

200

13,1

6654

230

Asi

a1,

188

1,14

071

4817

,108

350

83E

urop

e7,

669

05,

538

2,13

014

,012

1,54

81,

148

Lat

in A

mer

ica

530

425

418

016

,000

1,06

31,

000

n.a.

Not

app

licab

le.

Not

e: V

alue

s ar

e ba

sed

on 4

1 re

tren

chm

ent

prog

ram

s in

37

coun

trie

s. T

he p

erce

ntag

es c

ompu

ted

from

thi

s ta

ble

(for

exa

mpl

e, s

ever

ance

as

a pe

rcen

tage

of

tota

l cos

t) m

ay n

ot m

atch

the

per

cent

ages

giv

en in

tab

le 5

tha

t ar

e ba

sed

on a

cou

nt o

f pr

ogra

ms.

Sour

ce: A

utho

rs’ c

alcu

lati

ons

base

d on

indi

vidu

al c

ount

ry s

umm

arie

s (s

ee a

ppen

dix

tabl

es f

or c

ross

-cou

ntry

tab

ulat

ions

).


A wide variety of circumstances led to the retrenchment programs across coun-tries. The most prominent factors leading to retrenchment were fiscal crises anda general effort to reduce the size of government in the economy. However, insome cases, the compelling factors appeared to be structural problems with thetype and mix of government workers. Wage compression among public sectorworkers leading to morale and staffing problems was a common complaint.Overstaffing, including the problem of ghost workers, was another commoncomplaint. Finally, the downsizing of the military played a prominent role aswell. Not surprisingly, in many programs multiple factors led to the retrenchment.

The differences in the factors generating retrenchment are closely linked tothe mix of packages used across countries and continents. For some countries,the retrenchment episode was viewed as a one-time event correcting a perceivedrelatively narrow problem in a particular public sector agency or enterprise (forexample, ghost workers or productivity and morale problems due to wage com-pression). These one-time-event cases typically used some form of direct com-pensation (severance and enhanced pension) and accordingly were more likelyto use voluntary methods of retrenchment. By contrast, in some countries (forexample, the transition economies), the public sector retrenchment was part of afundamental change in the role of the public sector in the economy. In thesecases, the programs typically paid much greater attention to institutional changes(unemployment benefits, relocation assistance, training assistance) that acted asthe safety net for worker reallocation.

Much of the detailed information collected involves the method used to re-duce employment, the nature of the compensation provided, and the degree oftargeting. Tables 3, 4, and 5 provide summary information on these characteris-tics. Many programs (but less than half) used both involuntary and voluntaryreduction methods; relatively few programs used all three methods (table 3).Although voluntary programs appear to be the most popular in terms of the

Table 3. Distribution of Employment Reduction Methods in RetrenchmentPrograms, 1990s

Percentage PercentageType of method of programs of workers

Involuntary-hard (layoffs) 41.5 47.0Involuntary-soft (enforcement of rules,

removal of ghost workers) 65.0 30.0Involuntary (removal of ghost workers) 22.5 3.3Voluntary 77.5 23.0Both voluntary and involuntary 42.5 19.2All methods 17.5 13.4

Note: Values are based on 41 retrenchment programs in 37 countries. The percentage of workersmay differ from those computable from table 1 on account of missing values that are subsumed in thetotals in table 1. Missing values are excluded here.

Source: Authors’ calculations based on individual country summaries (see appendix tables for cross-country tabulations).


percentage of programs with a voluntary component, most reductions in em-ployment were achieved through involuntary methods. The inclusion of Hun-gary in our 41 programs has a clear influence on employment-weighted results.For the most part, we present unweighted results so that outliers such as Hun-gary do not play a disproportionate role. The removal of ghost workers playeda relatively minor supporting role in the employment reductions but was quiteprevalent in the African countries in our survey.

Table 4 indicates that the primary form of direct compensation to workerswas a severance payment. Relatively few programs offered no direct compensa-tion. Most programs involved some enhancement of the overall safety net in-tended to assist unemployed workers and workers attempting to relocate. Animportant component of the safety net enhancement was often the stipulation ofsome form of training assistance.

As discussed in section II, the underlying theory suggests that designing aprogram that targets specific individuals or groups of individuals is likely to beimportant to avoid adverse selection and to promote favorable selection. Weattempt to summarize the many different approaches to targeting used in prac-tice by classifying programs into three admittedly crude groups: skill-biased,age-biased, and neutral. Skill-biased programs restrict the program along de-tailed occupational or skill groupings. For example, the restructuring of the workforce in the tax administration in Peru was based on new and continuing work-ers passing a written test. Age-biased programs focus on voluntary retirement

Table 4. Distribution of Compensation and Transition Assistancein Retrenchment Programs, 1990s

Type of assistance Percentage of programs

Severance payment 68.3Pension enhancement 29.3No direct compensation 14.6Safety net 63.4Safety net (training) 53.7Unknown 12.2

Note: Values are based on 41 retrenchment programs in 37 countries.Source: Authors’ calculations based on individual country summaries (see appendix tables for cross-

country tabulations).

Table 5. Distribution of Targeting in Retrenchment Programs, 1990sType of targeting Percentage of programs

Skill-biased 53.7Age-biased 51.2Neutral 19.5Unknown 2.4

Note: Values are based on 41 retrenchment programs in 37 countries.Source: Authors’ calculations based on individual country summaries (see appendix tables for cross-

country tabulations).


(and associated pension enhancements) as the means for retrenchment. Neutralprograms offer a simple, common package to a wide class of workers with littleor no attempt to target specific groups. Some programs are classified as bothskill-biased and age-biased if they use a mix of packages that induce selection onboth types of criteria.

Table 5 indicates that targeting on the basis of skills or age (in the latter caseprimarily through early retirement programs) was important, but not pervasive.Almost 20 percent of the programs did not use targeting of workers. As we willsee in the following sections, the nature of targeting is closely related to a varietyof measures of success of the program.

Rehires, New Hires, and Other Program Characteristics

Tailoring a program to individual characteristics using some form of target-ing is likely to be important to avoid losing the most productive or essentialworkers. The absence of targeting may yield excessive losses of workers in keyareas or with key skills and thus will necessitate either rehiring some of the sameworkers who departed (a sure sign of poor targeting) or hiring new workers. Inmany programs, workers are rehired by the same organization and unit. In somecases, they are rehired by a different branch of the public sector. Using the infor-mation collected, we examine the simple bivariate relationships between the like-lihood of rehires and new hires and other program characteristics, including thenature of targeting.

Table 6 examines the link between rehiring and program characteristics. Themost striking results involve the role of targeting and the type of compensationoffered. Programs with rehiring were much less likely to use skill or age target-ing. For example, only 25 percent of programs with rehiring also used skill tar-geting compared with more than 60 percent of programs without rehiring. Theflip side of this is that more than 60 percent of the programs with rehiring did notargeting compared with less than 10 percent of programs without rehiring.These findings are consistent with the view that programs that fail to target yieldsevere adverse selection problems, with the most critical workers departing andcreating the subsequent need to rehire these same workers.

Some other interesting patterns emerge that are large in magnitude but notstatistically significant at conventional levels given the relatively small samplesize of 41 observations on individual programs. Specifically, programs with re-hiring were also less likely to have a safety net component and, in particular,some type of training component. The potential inference here is that the pres-sure to rehire may be greater if efforts are not made to assist workers in thetransition to the private sector.

Table 7 examines the analogous relationships between new hires and pro-gram characteristics. The largest and most significant results again involve thenature of targeting and the type of compensation assistance. Programs with newhires are all age-biased with associated pension enhancements and do not in-clude safety net components. It is not clear, of course, that evidence of new


hiring should be interpreted or treated in the same manner as rehiring. Rehiringthe same workers who departed is much more likely to indicate problems withthe program, while hiring new workers may involve some intended restructur-ing of the work force. Rehiring the same workers who were separated could bepart of a coherent plan of restructuring. For example, it might be difficult politi-cally to target specific workers for retrenchment. Given constraints on the abil-ity to target workers, the optimal strategy might be to induce separations enmasse and then to rehire specific workers. We found no evidence for such pur-posive rehiring. The results on age targeting and pension enhancements suggestthat for some programs the intent was to replace older, incumbent workers withnew workers.

Summary Financial Indicators

We calculate a summary financial indicator that measures the simple, finan-cial break-even period reflecting the number of years required before a programbreaks even on financial costs and benefits. Financial costs are the present dis-counted value of the compensation package or safety net expenses incurred.

Table 6. Relationship between Rehiring and the Characteristicsof Retrenchment Programs, 1990s

Program Percentage of programs Percentage ofcharacteristic with rehiring other programs

Employment reduction methodInvoluntary-hard 50.0 39.4Involuntary-soft 62.5 65.6Voluntary 62.5 81.3All methods 12.5 18.8

Nature of targetingSkill-biased 25.0* 60.6Age-biased 12.5** 60.6Neutral 62.5** 9.1

Compensation or assistanceSeverance payment 75.0 66.7Pension enhancement 37.5 27.3No direct compensation 12.5 15.2Safety net 50.0 66.7Safety net (training) 37.5 57.6

* The difference between programs with and without rehiring is statistically significant at the 10percent level.

** The difference between programs with and without rehiring is statistically significant at the 5percent level.

Note: Values are based on 41 retrenchment programs in 37 countries. Each value is the percentageof programs with or without rehiring that also has the indicated program characteristic.



This present discounted value is, in principle, calculated over an infinite hori-zon. In practice, for many of the cases, the financial costs are front-loaded withseverance payments, making this present discounted value easy to calculate. How-ever, when costs are in the form of a continuing safety net or pension liabilities,we generate appropriate present discounted value measures of the financial costs.(For some programs, the evaluation of the program by World Bank staff hadalready yielded such calculations.) Financial benefits are measured as the presentdiscounted value of the wage bill savings from the retrenched workers. The an-nual wage bill savings are assumed to be constant over time and equal to thewage bill savings relevant at the time of implementation of the program. Theseassumptions permit direct calculation of the break-even period as the number ofyears for the present discounted value of wage bill savings to equal the overallpresent discounted value of financial costs.

To achieve comparability of measures across countries, we use a commondiscount rate of 10 percent. Under these assumptions, the break-even period is ascale-free financial indicator that permits comparing and contrasting the finan-cial costs and benefits across programs. Information was sufficient to calculatethe break-even measure for about 40 percent of the programs.

Table 7. Relationship between New Hires and the Characteristicsof Retrenchment Programs, 1990s

Program Percentage of programs Percentage of programscharacteristic with new hires without new hires

Employment reduction methodInvoluntary-hard 0** 47.2Involuntary-soft 80 62.9Voluntary 80 77.1All methods 0 20.0

Nature of targetingSkill-biased 40 55.6Age-biased 100** 44.4Neutral 0 22.2

Compensation or assistanceSeverance payment 60 69.4Pension enhancement 60 25.0No direct compensation 0 16.7Safety net 0** 72.2Safety net (training) 0** 61.1

** The difference between programs with and without new hires is statistically significant at the 5percent level.

Note: Values are based on 41 retrenchment programs in 37 countries. Each value is the percentageof programs with or without new hires that also has the indicated program characteristic.



Across the programs surveyed, 22 percent yielded net losses, essentially im-plying an infinite break-even period. The presence of net losses reflects in partthe impact of rehires or new hires or of rising compensation for retained work-ers. Accordingly, some programs (for example, the Peruvian tax administrationauthority, SUNAT) yielded no wage bill savings as a result of rehires, new hires, oran increase in compensation. For the programs for which we can calculate afinite break-even period, the average break-even period is 2.27 years with a medianof 1.82 years, a maximum of 10 years, and a minimum of 0 years. Programswith immediate break-even periods (0 years) are those with involuntary reduc-tions without direct compensation or other assistance programs.

This simple financial indicator demonstrates the financial viability of the re-trenchment programs. Many of the programs yielded relatively rapid financialpayoffs. However, as discussed in section II, we would ideally like to quantifythe present discounted value of all costs and benefits from retrenching publicsector workers and evaluate programs accordingly. Where does this simple fi-nancial indicator fit into such a calculation? To address this question, it is usefulto return to inequality 3. For purposes of discussion, suppose we treat the lossfunction in inequality 3 as a simple linear function such that a dollar of expendi-tures on public sector programs yields a dollar in distortionary losses. Then,inequality 3 can be written as

(5) Wpublic – COMP + Pprivate – Csocial – Ppublic > 0.

Thus the measure of the break-even period essentially provides an evaluationof the program based on only the first two terms in inequality 5. A comprehen-sive evaluation, even in these crude terms, requires calculating all of the terms ininequality 5.

Unfortunately, data limitations imply that we cannot measure all of the com-ponents in inequality 5, and thus we cannot literally evaluate programs on thisbasis. However, for some programs we can go a few steps further. In particular,for some programs we can estimate the wages that retrenched workers receive inthe private sector. The assumptions that workers are paid their marginal prod-uct in the private sector, ignoring adjustment costs, and that productivity of theredundant (retrenched) workers is zero in the public sector permit a crude imple-mentation of the terms in inequality 5 using this additional information. Wherepossible, we calculate a modified break-even measure (denoted for labeling pur-poses as the economic payback period) using this additional information. Theeconomic payback period is the number of years required for the benefits fromthe reduced public sector wage bill to be such that the present discounted valueof costs equals the present discounted value of benefits. For this calculation, thecosts and benefits for all components other than the public sector wage bill arecalculated over an infinite horizon. In addition to measuring benefits based onretrenched workers’ earnings in the private sector, for programs with measur-able increases in productivity in the public sector (for example, SUNAT in Peru),we include such effects in the benefits. The average calculated economic pay-


back period is 2.1 years, and the median is 0.8 year across the programs forwhich we can generate the measure. (Sufficient information is available to calcu-late the payback period indicator for 20 percent of the programs.)

We calculate comparable summary financial and related indicators acrossprograms so that we can relate them to other program characteristics. Table 8characterizes the relationship between programs with net financial losses andother program characteristics. Programs with a net financial loss are much morelikely to include a voluntary component than programs without financial losses.Interestingly, programs with net financial losses are more likely to use targeting,particularly age targeting through pension enhancements. This result indicatesthat targeting may be expensive, at least measured in simple, financial terms.Table 9 provides a related look at the relationship between the magnitude of thebreak-even period and other program characteristics. Given that these are atbest crudely calculated, we divide programs into two groups—those in the lowerand upper tail of the distribution relative to the median break-even period (whichis 1.82 years). The results in table 9 reinforce those in table 8. Programs withrelatively high break-even periods are more likely to involve a voluntary compo-nent, more likely to use targeting, and more likely to use direct compensation(through either severance payments or pension enhancements).

Table 8. Relationship between Net Losses and the Characteristicsof Retrenchment Programs, 1990s

Program Percentage of programs Percentage ofcharacteristic with net losses other programs

Employment reduction methodInvoluntary-hard 44.4 40.6Involuntary-soft 50.0 68.8Voluntary 87.5 75.0All methods 25.0 15.6

Nature of targetingSkill-biased 66.7 50.0Age-biased 88.9** 40.6Neutral 0.0* 25.0

Compensation or assistanceSeverance payment 66.7 68.8Pension enhancement 55.6** 21.9No direct compensation 22.2 12.5Safety net 55.6 65.6Safety net (training) 55.6 53.1

* The difference between programs with and without net losses is statistically significant at the 10percent level.

** The difference between programs with and without net losses is statistically significant at the 5percent level.

Note: Values are based on 41 retrenchment programs in 37 countries. Each value is the percentageof programs with or without net losses that also has the indicated program characteristic.



Some of these patterns change substantially when we examine the distribu-tion of the payback period. As seen in table 10, once we incorporate even crudeinformation about productivity gains into the calculation, we no longer findthat targeting implies a high payback period. Not surprisingly, we still find thatprograms with high calculated payback periods are more likely to involve directcompensation and in particular severance pay.

To sum up, simple calculations of financial break-even periods indicate thatmany of the programs had relatively rapid financial payoffs. By contrast, a non-trivial fraction were financial losers. However, caution must be used in inter-preting such financial indicators as measures of the relative success of programs.There are many relevant components of the private and social costs and benefitsnot captured by these measures. Even the limited attempt made to incorporaterelevant economic costs and benefits indicates that inferences based on narrowlydefined financial indicators may be quite misleading.

Highlights of Individual Programs

We summarize the experiences of a select set of individual programs withthree objectives in mind. First, consideration of individual programs provides a

Table 9. Relationship between Financial Break-Even Periodsand the Characteristics of Retrenchment Programs, 1990s

Percentage of programs Percentage of programsProgram with below-median with above-mediancharacteristic break-even periods break-even periods

Employment reduction methodInvoluntary-hard 50.0 33.3Involuntary-soft 66.7 70.6Voluntary 33.3** 94.1All methods 0.0 23.5

Nature of targetingSkill-biased 0.0** 66.7Age-biased 16.7** 66.7Neutral 66.7** 5.6

Compensation or assistanceSeverance payment 66.7 83.3Pension enhancement 16.7 38.9No direct compensation 33.3 11.1Safety net 66.7 50.0Safety net (training) 50.0 50.0

** The difference between programs with break-even periods below and above the median value isstatistically significant at the 5 percent level.

Note: Values are based on 41 retrenchment programs in 37 countries. Each value is the percentageof programs with a break-even period below or above the median value that also has the indicatedprogram characteristic.



more comprehensive view of the scope and heterogeneity across programs. Sec-ond, the discussion of individual programs permits illustration of some of thekey patterns across programs. Third, our cross-country appendix tables and re-lated analyses admittedly leave out a lot. Part of the reason for this is lack ofdata suitable for quantifying effects in a summary fashion. For example, quanti-fying the nature of productivity gains associated with retrenchment as well asthe adjustment costs is difficult both in principle and in practice, given the avail-able data. In general, measuring productivity in the service sector, especially inthe public sector, is difficult; most work uses quantitative or technical measures(see Griliches 1992). However, some cases do provide qualitative informationabout these issues.

Here we highlight programs in Peru, Argentina, Uganda, Ghana, India, andHungary. Key characteristics are presented in table 11. In the discussion, theordering of the country/program cases reflects a rough attempt to select casesthat highlight the role of targeting, productivity gains, the worker safety net,and labor market adjustment. This ordering is only approximate because all ofthese issues are present for every program. In table 11, entries marked with a

Table 10. Relationship between Payback Periods and the Characteristicsof Retrenchment Programs, 1990s

Percentage of programs Percentage of programsProgram with below-median with above-mediancharacteristic payback periods payback periods

Employment reduction methodInvoluntary-hard 60 18.2Involuntary-soft 80 81.8Voluntary 20** 100.0All methods 0 18.2

Nature of targetingSkill-biased 20 54.5Age-biased 40 36.4Neutral 40 27.3

Compensation or assistanceSeverance payment 60** 100.0Pension enhancement 40 18.2No direct compensation 40** 0.0Safety net 60 45.5Safety net (training) 60 36.4

** The difference between programs with payback periods below and above the median value isstatistically significant at the 5 percent level.

Note: Values are based on 41 retrenchment programs in 37 countries. Each value is the percentageof programs with a payback period below or above the median value that also has the indicated programcharacteristic.


43

Tab

le 1

1. K

ey C

hara

cter

isti

cs in

Sel

ecte

d R

etre

nchm

ent

Pro

gram

s, 1

990s

Arg

enti

naP

eru

Uga

nda

Fede

ral

Civ

ilC

ivil

Cha

ract

eris

tic

gove

rnm

ent

Rai

lroa

dsG

hana

Hun

gary

Indi

ase

rvic

eSU

NA

Ta

serv

ice

Mili

tary

Tar

geti

ngSk

ill-b

iase

dN

oN

oY

esY

esN

oN

o*Y

es*

Yes

*Y

esA

ge-b

iase

dY

esY

esY

esY

esY

esN

o*Y

es*

Yes

*N

oN

eutr

al—

——

——

Yes

*—

*—

—

Red

ucti

on m

etho

dIn

volu

ntar

y-ha

rdN

oN

oN

oY

esN

oY

es*

No*

Yes

*Y

esIn

volu

ntar

y-so

ftY

esY

esY

es—

No

Yes

*Y

es*

Yes

*Y

esV

olun

tary

No

Yes

Yes

—Y

esY

es*

Yes

*Y

es*

Yes

Reh

ires

No

No

No

No

No

Yes

*N

o*N

o*N

oN

ew h

ires

Yes

No

No

No

No

No*

Yes

*N

o*N

o

Fina

ncia

l ind

icat

ors

Bre

ak-e

ven

peri

od0.

41*

1.56

*1.

82*

Net

loss

Net

loss

2.6*

Net

loss

*N

et lo

ss*

2.7

Payb

ack

peri

od—

—1.

66*

——

—*

0.00

23*

—1.

2

Pro

duct

ivit

y ga

ins

Org

aniz

atio

nal

Yes

Yes

No

——

—*

Yes

—N

o(q

uant

itat

ive)

(qua

ntit

ativ

e)(m

onet

ary)

Wor

ker

——

Yes

——

—*

——

Yes

*(m

onet

ary)

(mon

etar

y)

Safe

ty n

et p

rovi

sion

No

No

Yes

Yes

*Y

es*

No

*N

oY

esY

es(m

ost

wor

kers

(mos

t w

orke

rs r

eem

ploy

ed)

reem

ploy

ed)

— N

ot a

vaila

ble.

* A

key

issu

e th

at m

otiv

ates

our

incl

usio

n of

the

pro

gram

for

spe

cial

dis

cuss

ion.

a. T

he P

eru

tax

adm

inis

trat

ion

auth

orit

y.So

urce

: Aut

hors

’ cal

cula

tion

s ba

sed

on in

divi

dual

cou

ntry

sum

mar

ies

(see

app

endi

x ta

bles

for

cro

ss-c

ount

ry t

abul

atio

ns).

**

*

**


star are key issues that motivate our inclusion of the program for specialdiscussion.

PERU. Two programs were initiated as part of a broader exercise of fiscalausterity and adjustment in Peru. Problems included the underfunded pensionsystem, the posting of ghost employees by the regions to enhance employment-based federal transfers, and the erosion in public salaries, making it difficult tohire and retain qualified workers.

With the support of external donors, Peru initiated two labor adjustment pro-grams in 1991, one for the civil service and one for SUNAT. These were completedin 1993 and 1992, respectively (World Bank 1994b). The civil service programused all three involuntary and voluntary employment reduction methods, sepa-rating about 250,000 workers over three years. Induced departures used bothlump-sum severance and pension enhancements. We found little evidence of tar-geting by skill. Targeting by age was implicit in the use of pension enhancementsto induce voluntary separation. Poor targeting aggravated the shortage of humanresources, with many of the most qualified staff leaving. The poor targeting andaccompanying shortages probably led to the significant rehiring of separatedworkers. Further, federal government staff reductions were offset by increases inemployment by regional governments, mostly by rehiring erstwhile federal staff.In total, 163,000 of the originally retrenched workers were rehired. Severancepackages of about $1,000 were provided to less than half (112,000) of the work-ers separated. This limited the direct financial losses associated with the signifi-cant rehiring. Our simple measure of the break-even period for this program is2.6 years, more than the median of 1.82 years for all the programs studied. How-ever, this measure does not reflect the loss of productivity associated with shuf-fling the same workers in and out of the same positions.

In contrast, the other program in Peru appears to have been a model of goodtargeting. The SUNAT (tax authority) program also used a mix of voluntary andinvoluntary reduction methods. Voluntary separations came with an enhancedpension. Involuntary workers were selected on the basis of a written test. Thustargeting was worker-specific and objectively determined. Two-thirds of the workforce (2,034 workers) was separated. Subsequently, SUNAT hired 1,309 new work-ers, again based on a written test. Because SUNAT established objective levels ofproductivity and competence, little basis remained for rehiring (skilled butseverance-induced) separated workers. Rehiring was barred for 10 years, andnone was found. The average salary for affected workers increased from $50 permonth to $1,000 per month. Deflation of wages using the consumer price indexinstead of the exchange rate produces an increase of similar magnitude; the in-stantaneous increase drops from 20 to 18.6 times. Tax collections more thandoubled, and so did SUNAT’s revenues (2 percent of tax collections). These wereinsufficient to cover the salary increase, and the scheme incurred a net financialloss of $47 million in present value terms. However, the entire increase andimprovement in tax collections can be interpreted as a substantial gain for the


government (and, in principle, the economy).3 Incorporating the tax collectiongains into our calculation of the payback period yields a payback period of only0.002 year. This highlights the importance of evaluating program performancealong multiple dimensions, evaluating not just organization-level financial costsand benefits but also the broader impact of the program. Neither of these twoprograms emphasized job search, relocation, training, or other forms of assis-tance to reduce the adjustment costs for the retrenched workers.

ARGENTINA. With external assistance from the World Bank, Argentina under-took retrenchment programs for the federal administration and the railroads (aspart of a broader program for public enterprises) during 1990–92 and 1991–94,respectively. In a case study on public sector retrenchment in Argentina since1990, Robbins (1996) examines issues of adverse selection and post-retrenchmentworker status using micro data. Given chronic public sector deficits and endemicinflation, extensive public sector reform was considered essential. Wagesaccounted for 70 percent of federal expenditure, excluding interest paymentsand transfers. The level of employee redundancy was estimated at 50 percent(World Bank 1995a). For this reason, staff restructuring was an importantcomponent of overall reform.

The federal administration program retrenched 400,000 workers. No explicittargeting by worker is found. Involuntary separations implied some implicit tar-geting in the selection of redundant workers. Some evidence indicates an agebias. Also, targeting by function was attempted in that the federal government’srole was restricted to the provision of security, social services, and economicmanagement. Rehiring was barred by law, and none is found, despite the largescale of retrenchment and the absence of targeting. We find no evidence of rehir-ing in the source reports (World Bank 1994d, 1995a, 1995c); however, someanecdotal evidence indicates the occurrence of rehiring.

The railroads program separated about 73,000 workers using both voluntaryand involuntary reduction methods. An important feature of Argentinean pro-grams is the heterogeneity in compensation amounts. In federal administration,the average severance payment per worker was $3,000, amounting to a one-time cost of $425 million (World Bank 1995a). In railways, the average sever-ance per worker was $12,000, or a one-shot payment of $360 million (WorldBank 1995c). Savings in the annual wage bill was expected to be $1 billion in thefederal service and $238 million in the railroads. The calculated break-even pe-riods are 0.4 year and 1.6 years, respectively, in the federal administration andthe railroads, with the difference reflecting the higher compensation per workerin railroads.

Although difficult to measure, restructuring in both railways and federal ad-ministration apparently was characterized by substantial productivity gains.

3. This discussion neglects the welfare costs (for example, reduction in hours and output) due toincreased enforcement of tax collections.


Railways experienced an increase in freight miles per worker and passenger milesper worker of about six to seven times. Federal administration experienced a 50percent cut in processing time.

UGANDA. In Uganda, high fiscal deficits combined with inadequate pay levelsand wage compression led to a civil service labor restructuring program in 1992–94. Concomitantly, the military reduced its size in line with peacetime nationalpriorities and conducted a labor restructuring program in 1992–95. Both werepartly supported by external financing. The civil service program employed allthree reduction methods, restricting the package offer to a fraction of the workersseparated. Separations were targeted by skill and tenure. Some critical jobs andpersonnel were “ring-fenced,” that is, protected from retrenchment. Such workersdid not have the option of obtaining enhanced benefits associated with voluntaryseparation.

The gross number of separations was approximately 150,000. The averagecompensation per worker was $320, or about twice the annual gross domesticproduct (GDP) per capita (Uganda Ministry of Public Service 1994). This totalexpense of $16 million constituted a measured net financial loss because therewas no saving on the wage bill. The wage bill increased due to salary revisionsfor the remaining civil service staff and was intended to improve service perfor-mance. Interviews with program advisers revealed that the salary hike was in-tended to achieve parity between the civil service average wage and the livingwage in Uganda so as to motivate the civil service staff and generate productiv-ity increases. Although an evaluation of the civil service staff performance wasnot available at the time of our data collection, measurement indicators to evaluatethe performance improvements were planned by the Economic DevelopmentInstitute of the World Bank.

The military program was a postwar retrenchment exercise. It used both vol-untary and involuntary reduction methods to reduce strength by 44,000 em-ployees. Skill targeting was based on soldier performance and military require-ments. Average compensation was $955, provided partly in kind and distributedover several months. The calculated break-even period is 2.7 years. In Uganda,some measurement of post-retrenchment earnings for veterans is available. Ac-tual post-separation earnings for the military programs were up to 71 percent ofGDP per capita (World Bank 1995b). Including these earnings of a total $5.4million per year in Uganda’s case yields a calculated payback period of 1.2 years.The rarely observed actual data on post-separation earnings are interesting, par-ticularly because the earnings were lower than the average. Although this evi-dence is too idiosyncratic to draw general inferences, it highlights the difficultyin generating the appropriate cost-benefit calculations. Using average earningsin the private sector as a proxy (even when it is available) is obviously inappro-priate in this case.

An interesting feature of the military program was the safety net nature of thecompensation, which enabled veterans to reintegrate into civilian society. Assis-


tance in kind was tailored to the intended post-separation occupation; for in-stance, farmers were provided land and farm implements. Furthermore, the as-sistance was known to be temporary (six months), creating an incentive for theveteran to hasten adjustment and thereby limiting program costs.

GHANA. Motivated by a very large public sector wage bill as a fraction of totalgovernment expenditures, the government of Ghana initiated a program for thecivil service in 1987 with partial financing from the World Bank. The programconcluded in 1992. It was driven by voluntary departures induced by additionalseverance and complemented by involuntary-soft measures—removing ghostworkers and enforcing retirement age. Voluntary departure, with compensation,was conditional on employment being noncritical to the performance of the unit.The government thus sought to protect necessary skills. More than three-fourthsof the total 73,000 workers separated received the package. Average compensationwas $700. The formula of two months of base pay for each year of uninterruptedservice exceeded the legally required four months of base pay in most cases.Anecdotal evidence indicated some rehiring in subvented organizations (attachedto ministries but not covered under the budgetary process). The calculated break-even period is 1.8 years. Despite a generous package offer, the break-even periodis relatively low.

The government planned extensive training through the Program of Actionto Mitigate the Social Costs of Adjustment (PAMSCAD) to assist workers mak-ing a transition to the private sector. However, assistance for job search andplacement and for retraining was lacking. PAMSCAD offered courses in entre-preneurial development and provided inputs, including land for potential farm-ers. In rural areas, retrenched workers were also eligible to participate in food-for-work schemes. A sample survey of the retrenched workers finds that about10 percent of them quit the labor force (Alderman, Canagarajah, and Younger1996). Of the rest, 97 percent were reemployed by the second year, about 20percent in formal sector wage jobs and the rest in self-employment or informalsector jobs.

INDIA. Fiscal and external payments deficits led India to initiate a stabilizationand structural adjustment program in 1991. A program to support public sectorlabor restructuring was required. This retrenchment support program was directedat public enterprises declared to be sick (with several years of accumulated losses).A voluntary retrenchment program in the sick public sector textile firms separatedabout 70,000 workers in 1993–94. The average cost per worker was about$17,000. The formula used was 30 days of wages for each year worked, comparedwith the legally required 15 days of wages for each year of permanent service.The scheme incurred a net loss of $276 million in present value terms, given theexorbitant compensation. The high compensation reflects in part the effects of arigid law against retrenchment and closure in India (see Basu, Fields, and Debgupta1996). Given the incentives, the voluntarily departing workers may have had


long tenures, increasing the severance amount granted. No explicit targetingmechanism by tenure or age is found. The maximum skill level among theretrenched workers was that of a supervisor.

Results from a sample survey indicate that all retrenched workers remained inthe labor force, and 80 percent were reemployed (PA Consulting Group 1993).Of the total surveyed, 32 percent were in wage jobs, one-fourth of them in thesame industry, that is, textiles. The rest (48 percent) were self-employed.

HUNGARY. In Hungary, reform and stabilization programs aimed at a transitionto a competitive economic system led to mass dismissals in state enterprises during1990–92. About 1.7 million workers constituting 8.7 percent of the country’slabor force were separated. This is extremely large compared with the scope ofother programs (0.2 percent of the labor force in Ghana and 0.01 percent inIndia). The annual wage saving amounted to $298 million but was exceeded bythe increase in the annual safety net expenditures of $858 million. The exercisewas clearly financially costly but reflected a fundamental change in the structureof the economy. A comprehensive set of institutions designed to act as a safetynet for worker reallocation is very common in Western economies, and Hungarywas obviously attempting to follow that model.

Surveys indicate that retrenched workers remained in the labor force (Com-mander and others 1994). Most of them were reemployed in the private sector,in trade and service industries. The average private manufacturing wage wasfound to be about 70 percent of the state firm wage (indicating the size of therent in public sector firms). The nature of the labor market adjustment appearedto change as the transition process proceeded and accelerated. Early on, many ofthe retrenched workers left voluntarily and left the labor force (for example,through voluntary retirement). Others left voluntarily and transited directly toother jobs. However, as involuntary layoffs increased, a larger fraction of theretrenched workers entered unemployment. Moreover, the available evidencesuggests that unemployed workers had an increasingly difficult time finding jobs.Early in the transition, in February 1991, the outflow rate from unemploymentimplied a steady-state duration of unemployment of 7 months. By November1992, the implied steady-state duration was 50 months.

Aggregate Factors

Another means of evaluating the programs is to look at relevant macroeco-nomic indicators prior to and during the program, including indicators of fiscalposture. Examining such macro indicators provides a rough independent checkof the factors that led to a retrenchment episode. Further, evaluating macroindicators prior to and during the program provides another means of evaluat-ing the impact of the program. In many cases, the actual retrenchment programis too small by itself to have important macro effects. Exceptions are the pro-grams in Eastern Europe (for example, Hungary) in which there was massivepublic sector downsizing. However, even in cases where the specific program


analyzed was too small by itself to have macro effects, the program may havebeen part of a larger effort to restructure and downsize the public sector. Theresults that follow can be interpreted in this light.

For each of the programs surveyed, we acquired data on the unemploymentrate, the real GDP growth rate, the ratio of the current budget deficit to GDP, theratio of domestic debt to GDP, the ratio of foreign debt to GDP, and the ratio ofgovernment spending to GDP. Several sources were used for this purpose, includ-ing the Penn World Tables (Summers and Heston 1991), the cross-country labormarket database generated by Rama (1995), World Bank (various issues), andInternational Monetary Fund (various issues). To remove the influence of idio-syncratic effects across countries, all variables are characterized in terms of de-viations from country-specific means. Country-specific means for each variableare calculated from data for 1980–92. The data are not complete for all vari-ables and all countries over this entire horizon. The unemployment rate series isprobably the worst in this regard. Considering deviations from country-specificmeans mitigates the problems with limited data availability. However, appro-priate caution should be used in considering these results (particularly for unem-ployment), with the additional usual caveats that considerable measurement er-ror is likely in the aggregates available by country. Table 12 presents the patternsof these variables for the five years prior to initiation of the retrenchment pro-gram in each country and during the program. For example, for a specific coun-try, if the retrenchment program began in 1990 and ended in 1992, the prioryear calculations reflect data for 1980–89 and the calculations during the pro-gram reflect data averaged over 1990–92.

Unemployment rises and GDP growth rates fall in the years leading up to aprogram (table 12). The unemployment rate appears to fall somewhat duringprograms, while GDP growth remains below average. These patterns are roughlyconsistent with the view that retrenchment programs are often precipitated byeconomic crises. The evidence on fiscal indicators is somewhat mixed at firstglance. The government share of GDP rises steadily prior to the onset of a pro-gram and falls during the program. Somewhat surprisingly, there is no accom-panying pattern for the deficit-to-GDP ratio or the debt-to-GDP ratio. For both ofthese domestic fiscal indicators, there is an improvement in the fiscal posturepreceding the program. However, in part this appears to reflect the rising for-eign debt-to-GDP ratio that precedes a program and continues during a program.Putting these pieces together is consistent with the view that these retrenchmentprograms are often part of a more general austerity program that is supported inpart by foreign assistance.

Missing Pieces: Measurement and Characterization of Adjustment Costs

For evaluation, one of the biggest missing pieces is associated with the con-ceptual and measurement problems involved in quantifying private and socialadjustment costs. Although quantification is beyond the scope of readily avail-able information, a few qualitative observations can be made regarding the cross-

50

Tab

le 1

2. C

ondi

tion

s be

fore

and

dur

ing

Ret

renc

hmen

t P

rogr

ams:

Dev

iati

ons

from

Cou

ntry

-Spe

cifi

c M

eans

, 199

0sFi

ve y

ears

Four

yea

rsT

hree

yea

rsT

wo

year

sO

ne y

ear

Dur

ing

Indi

cato

rbe

fore

pro

gram

befo

re p

rogr

ambe

fore

pro

gram

befo

re p

rogr

ambe

fore

pro

gram

prog

ram

Une

mpl

oym

ent

rate

–0.0

1–0

.26

–0.0

51.

090.

87–0

.18

Rea

l GD

P gr

owth

rat

e**

–0.0

50.

82–0

.70

–3.2

5–0

.41

–2.0

8D

efic

it/ G

DP *

*0.

012

–0.0

05–0

.008

–0.0

10–0

.012

–0.0

34D

omes

tic

debt

/GD

P **

–0.0

36–0

.049

–0.0

41–0

.036

–0.0

63–0

.129

Fore

ign

debt

/ GD

P–0

.050

–0.0

580.

026

0.05

50.

098

0.14

0G

over

nmen

t sp

endi

ng/G

DP

**0.

020

0.02

10.

021

0.01

40.

001

0.00

7

** I

ndic

ates

tha

t th

e F-

test

for

the

nul

l hyp

othe

sis

that

all

of t

he c

oeff

icie

nts

in t

he r

ow a

re e

qual

is r

ejec

ted

at t

he 5

per

cent

leve

l.N

ote:

Val

ues

are

base

d on

41

retr

ench

men

t pr

ogra

ms

in 3

7 co

untr

ies.

The

def

icit

num

bers

are

equ

al t

o ex

pend

itur

es le

ss r

even

ues

and

igno

re a

ny g

rant

s an

d/or

loan

s.So

urce

: Aut

hors

’ cal

cula

tion

s ba

sed

on a

var

iety

of s

ourc

es in

clud

ing

the

Penn

Wor

ld T

able

s (S

umm

ers

and

Hes

ton

1991

), W

orld

Tab

les

(Wor

ld B

ank,

var

ious

issu

es),

Int

erna

tion

al F

inan

ce S

tati

stic

s (I

nter

nati

onal

Mon

etar

y Fu

nd, v

ario

us is

sues

), a

nd R

ama

(199

5). S

ee t

ext

for

met

hodo

logy

.


country variation in labor market adjustment costs. Specifically, qualitative evi-dence suggests that it is important to consider the impact of the formal andinformal sectors. The discussion that follows on the formal and informal labormarket is inadequate on a number of dimensions and does not reflect the richbody of research that has been conducted on the role of formal, informal, andrural sectors for labor market dynamics in developing countries (see Mazumdar1989 for an overview of this research). Our point here is simply to highlight thepotential importance of these considerations for cross-country differences in thenature of labor market adjustment. In countries in which the informal sectorplays a large role (for example, Ghana), many of the retrenched public sectorworkers found employment relatively quickly, albeit rarely in the formal wagesector. By contrast, in the transition economies, retrenched workers from publicsector enterprises faced increasing difficulty in leaving unemployment becauseof difficulties in finding employment in the formal wage sector.

Stated very roughly, these observations suggest the following potentially im-portant difference across countries in the relevant labor market adjustment. Oneset of countries seems to yield relatively quick adjustment, with the informalsector absorbing retrenched public sector workers. Other countries seem to yieldlong adjustment, with retrenched workers from public sector enterprises experi-encing long spells of unemployment while seeking jobs in the formal privatesector. These differences are further reflected in the mix of policies observedacross countries. For example, in Africa and to some extent in Latin America,many of the programs surveyed involved direct compensation and relatively fewchanges in the formal worker safety net. By contrast, in transition economies,the resources for retrenched workers were devoted primarily to the worker safetynet.

This contrast is probably overstated in several respects. First, employment inthe informal sector may involve substantial underemployment. Second, workersreported to be officially unemployed are arguably often engaged in some unre-ported work in the informal sector. Nevertheless, even this rough characteriza-tion highlights the potentially important role of differences across countries inthe structure of economies and, in turn, the differences in the nature of labormarket adjustment costs.

We have learned a great deal from industrial economies (and some transitioneconomies) about labor market flexibility and adjustment by examining the flowof workers and jobs (see, for example, Davis, Haltiwanger, and Schuh 1996).For example, some industrial economies, including the United States and Canada,exhibit high rates of flow into and out of unemployment, while others, includingFrance and Italy, exhibit low rates (see OECD 1993, 1996). The variation in therates of flow into and out of unemployment translate into striking differences inthe importance of long-term unemployment across countries. In the United Statesthe percentage of long-term unemployed (more than 12 months) in the early1990s was about 6 percent of total workers, while the equivalent figure in Italywas 71 percent. Even this limited and indirect evidence yields two immediate


inferences for the analysis of public sector retrenchment. First, the low outflowrates from unemployment and the accompanying importance of long-term un-employment in some countries imply that the magnitude of the adjustment costsare potentially very large. Second, the variation across countries is substantial,implying that evaluation of a specific program in a specific country will dependcritically on the nature of the labor market adjustment in that country.

IV. CONCLUDING REMARKS

This survey and analysis of cross-country experiences with retrenchment re-flects the gathering and processing of a wealth of information about individualprograms. The nature of the compensation packages, the employment reductionmethods used, the nature of targeting, the worker safety net components of theprogram, the financial costs and benefits, and the nature of adjustment (includ-ing the presence of significant rehiring of the same workers) have been docu-mented and examined in detail. One of the primary objectives and thus contri-butions of this article has been to gather all of this information into one place inan easily accessible manner.

Several interesting findings emerge from analysis of the basic facts. Theorysuggests the importance of individually tailoring programs to account for workerheterogeneity and to avoid adverse selection problems. Consistent with this view,we found that programs were much less likely to exhibit problems with rehiringif they used targeting on the basis of skills and age, multiple methods of employ-ment reduction, and a combination of compensation packages that includedenhancements of the safety net for assisting the reallocation of workers. We alsofound that programs with this multidimensional approach and with targetingtended to be financially expensive. However, both quantitative and qualitativeinformation suggests that there is a potentially large payoff in productivity gainsand in lower adjustment costs.

Our analysis of the financial indicators revealed considerable heterogeneity infinancial viability across the programs. Many of the programs had a rapid finan-cial payoff in the sense that the wage bill savings from retrenchment quicklycovered the financial costs of the programs resulting from worker compensationand assistance. Alternatively, a nontrivial fraction of the programs were clearfinancial losers. However, we emphasize that although simple financial indica-tors are of obvious interest, they are inappropriate for evaluating the relativesuccess of programs because they omit many of the potentially relevant privateand social costs and benefits.

Unfortunately, many of the relevant private and social costs and benefits aredifficult to quantify, given data (and conceptual) limitations. Especially difficultto quantify are the individual and economywide (social) adjustment costs thatare incurred as part of restructuring programs. The data required to assess theseadjustment costs are generally not available, and this is accompanied by difficultconceptual issues. This problem has both micro and macro dimensions. On the


micro side, little effort has been made to collect data systematically on the post-separation experiences of retrenched workers. On the macro side, characteriz-ing and understanding the nature of these adjustment costs requires understand-ing of the myriad factors (for example, institutions) that affect the processes oflabor market adjustment within individual countries. The large observed differ-ences across even industrial economies provide prima facie evidence in supportof the need for this type of information. In countries with massive public sectorretrenchment episodes, the labor market adjustment process is endogenous tothe retrenchment program itself. Given the importance of the issues, more re-sources should be devoted to developing the data necessary to evaluate theseadjustment costs.

54

Tab

le A

-1.

Pub

lic S

ecto

r R

etre

nchm

ent,

by

Cou

ntry

and

Pro

gram

, 199

0s(n

umbe

r of

wor

kers

)B

y in

stru

men

t

Reg

ion,

Tim

eE

mpl

oym

ent

Invo

lunt

ary-

Invo

lunt

ary-

New

Num

ber

coun

try,

and

cas

epe

riod

redu

ctio

nha

rdso

ftV

olun

tary

Sepa

rati

ons

Reh

ires

hire

s

Afr

ica

1B

enin

Civ

il Se

rvic

e19

91–9

410

,061

1,04

05,

721

3,30

010

,061

00

2B

urki

na F

aso

Join

tR

ailw

ay L

ine

1994

a1,

585

——

—1,

585

00

3C

amer

oon

Civ

il Se

rvic

e19

89–9

46,

500

2,80

4—

3,69

66,

500

00

4C

ape

Ver

de P

ublic

Ent

erpr

ises

1992

–97

247

0—

247

247

00

5C

entr

al A

fric

an R

epub

licC

ivil

Serv

ice

1987

–90

1,10

00

2,27

572

53,

000

1,90

00

6C

ongo

Civ

il Se

rvic

e19

95a

8,00

0—

8,00

00

8,00

00

07

Eth

iopi

a M

ilita

ry19

91–9

554

1,20

0—

—0

547,

200

6,00

00

8G

hana

Civ

il Se

rvic

e19

87–9

273

,810

014

,000

59,8

1073

,810

00

9K

enya

Civ

il Se

rvic

e19

94–9

730

,800

—25

,000

5,80

030

,800

00

10M

alaw

i Civ

il Se

rvic

e19

94a

——

——

——

—11

Mau

rita

nia

Publ

ic E

nter

pris

es19

90–9

41,

900

——

—1,

900

00

12N

amib

ia M

ilita

ry19

91–9

449

,500

57,0

000

057

,000

7,50

00

13Se

nega

l Civ

il Se

rvic

e19

89–9

14,

357

01,

800

4,30

06,

100

01,

743

14Si

erra

Leo

ne C

ivil

Serv

ice

1992

–97

27,4

5220

,852

1,10

06,

000

27,9

5250

00

15U

gand

a C

ivil

Serv

ice

1992

–94

91,3

3911

,339

75,0

005,

000

91,3

390

016

Uga

nda

Mili

tary

1992

–95

44,2

1111

,000

8,00

09,

000

44,2

110

0

Asi

a17

Ban

glad

esh

Jute

Sec

tor

Publ

ic E

nter

pris

es19

94–9

522

,250

04,

000

21,2

5025

,250

3,00

00

18C

ambo

dia

Civ

il Se

rvic

e19

95/9

6b50

,000

——

—50

,000

00

19C

hina

She

nyan

g R

egio

nR

efor

m19

94–2

002

7,00

0—

7,00

00

7,00

00

020

Indi

a Pu

blic

Ent

erpr

ises

1993

–94

69,4

660

069

,466

69,4

660

0

55

21L

ao P

DR C

ivil

Serv

ice

1989

–93

21,6

00—

——

21,6

000

022

Paki

stan

Pub

lic E

nter

pris

es19

91–9

37,

495

00

7,49

57,

495

00

23Pa

kist

an S

indh

Reg

ion

Ref

orm

1993

–97

6,50

00

—6,

500

6,50

00

024

Sri L

anka

Civ

il Se

rvic

e19

91–9

249

,000

00

68,0

0068

,000

019

,000

Eur

ope

25A

lban

ia P

ublic

Ent

erpr

ises

1992

253,

000

253,

000

00

253,

000

00

26H

unga

ry P

ublic

Ent

erpr

ises

1990

–92

1,66

1,00

0—

——

1,66

1,00

00

027

Kaz

akhs

tan

Publ

ic E

nter

pris

es19

9317

2,95

940

,959

132,

000

—17

2,95

90

028

Mac

edon

ia P

ublic

Ent

erpr

ises

1988

–93

112,

000

42,0

000

70,0

0011

2,00

00

029

Pola

nd P

ublic

Ent

erpr

ises

1990

–93

547,

300

547,

300

00

547,

300

00

30R

ussi

an F

eder

atio

n C

oal

Sect

or19

96b

72,3

00—

——

72,3

000

031

Tur

key

Publ

ic E

nter

pris

es19

93–9

435

,000

0—

35,0

0035

,000

00

Lat

in A

mer

ica

32A

rgen

tina

Pub

lic E

nter

pris

es19

91–9

472

,818

042

,818

30,0

0072

,818

00

33A

rgen

tina

Fed

eral

Adm

inis

trat

ion

1990

–92

405,

995

042

4,09

50

424,

095

018

,100

34B

oliv

ia M

inin

g—Pu

blic

Cor

pora

tion

1991

–94

4,25

10

04,

599

4,59

934

80

35B

razi

l Civ

il Se

rvic

e19

90–9

10

100,

000

00

100,

000

100,

000

036

Chi

le C

ivil

Serv

ice

and

Para

stat

alO

rgan

izat

ions

1973

–77

91,1

0091

,100

00

91,1

000

037

Col

ombi

a T

ouri

sm a

nd T

rans

port

Min

istr

y19

90–9

212

,000

00

12,0

0012

,000

00

38E

cuad

or C

ivil

Serv

ice

1992

–94

40,0

000

040

,000

40,0

000

039

Mex

ico

SOC

EFI

—M

inis

try

of T

rade

and

Indu

stry

1989

–92

4,15

00

2,00

03,

000

5,00

00

850

40Pe

ru C

ivil

Serv

ice

1991

–93

100,

595

0—

112,

000

263,

654

163,

059

041

Peru

SU

NA

T—

Tax

Col

lect

ing

Aut

hori

ty19

91–9

272

50

——

2,03

40

1,30

9

— N

ot a

vaila

ble.

Not

e: I

n se

vera

l cas

es, t

he n

umbe

r of

wor

kers

sep

arat

ed b

y in

stru

men

t do

es n

ot a

dd u

p to

tot

al s

epar

atio

ns, o

win

g to

par

tial

ava

ilabi

lity

of in

form

atio

n.a.

Ong

oing

.b.

Pro

pose

d.So

urce

: A

utho

rs’ c

alcu

lati

ons.


Table A-2. The Financial Costs of Public Sector Retrenchment, by Countryand Program, 1990s

Financial costs(millions of dollars) Cost per

Region, Total Severance Enhanced Safety workerNumber country, and case cost payments pension net (dollars)a

Africa1 Benin Civil Service 21 21 0 0 6,4242 Burkina Faso Joint

Railway Line — — — — —3 Cameroon Civil Service 7 7 0 — 1,9974 Cape Verde Public

Enterprises 3 2 0 0 10,2605 Central African Republic

Civil Service — — — — —6 Congo Civil Service — — — — —7 Ethiopia Military 200 0 0 200 365b

8 Ghana Civil Service 42 42 0 — 7009 Kenya Civil Service 20 0 20 — 3,44810 Malawi Civil Service 20 20 0 0 —11 Mauritania Public

Enterprises 9 9 0 — 4,910c

12 Namibia Military 38 13 — 25 658b

13 Senegal Civil Service 80 80 — 0 13,166c

14 Sierra Leone Civil Service 2 2 0 0 35315 Uganda Civil Service 16 16 — — 320d

16 Uganda Military 42 42 0 — 955c

Asia17 Bangladesh Jute Sector

Public Enterprises 56 56 0 — 2,62118 Cambodia Civil Service 50 50 0 — 1,000c

19 China Shenyang RegionReform 0 0 0 — 0

20 India Public Enterprises 1,188 1,140 — 48 17,10821 Lao PDR Civil Service 10 10 — — 470c

22 Pakistan Public Enterprises 25 25 0 — 3,31823 Pakistan Sindh Region

Reform — — — — —24 Sri Lanka Civil Service 71 0 71 0 1,040

Europe25 Albania Public Enterprises 6 0 0 6 24b

26 Hungary Public Enterprises 859 0 0 859 517b

27 Kazakhstan PublicEnterprises — — — — —

28 Macedonia PublicEnterprises 50 0 50 — 714

29 Poland Public Enterprises 7,669 0 5,538 2,130 14,012b

30 Russian FederationCoal Sector — — — — —

31 Turkey Public Enterprises — — — — —


Latin America32 Argentina Public

Enterprises 360 360 0 0 12,00033 Argentina Federal

Administration 425 425 0 0 1,002b

34 Bolivia Mining—PublicCorporation 74 74 0 0 16,000

35 Brazil Civil Service — — — — —36 Chile Civil Service and

Parastatal Organizations — 0 — — 037 Colombia Tourism and

Transport Ministry — — — — —38 Ecuador Civil Service 200 200 0 — 5,00039 Mexico SOCEFI—Ministry

of Trade and Industry — — — — —40 Peru Civil Service 530 112 418 0 4,73541 Peru SUNAT—Tax

Collecting Authority 2 1 1 0 1,131c

— Not available.a. Amount of severance per (voluntarily retrenched) worker.b. Cost per worker; workers separated involuntarily.c. Average severance amount per worker (for all retrenched workers).d. Average severance using all workers (excluding the 42,000 ghost workers).Source: Authors’ calculations.

Table A-2 (continued)Financial costs

(millions of dollars) Cost perRegion, Total Severance Enhanced Safety worker

Number country, and case cost payments pension net (dollars)a

58

Tab

le A

-3.

The

Fin

anci

al B

enef

its

of P

ublic

Sec

tor

Ret

renc

hmen

t, b

y C

ount

ry a

nd P

rogr

am, 1

990s

(mill

ions

of

dolla

rs)

Reg

ion,

Wag

e bi

llW

ages

Num

ber

coun

try,

and

cas

eT

otal

savi

ngs

(ann

ual)

Sepa

rati

ons

Reh

ires

New

hir

esO

ther

a

Afr

ica

1B

enin

Civ

il Se

rvic

e0

0—

——

—2

Bur

kina

Fas

o Jo

int

Rai

lway

Lin

e—

——

——

—3

Cam

eroo

n C

ivil

Serv

ice

——

——

——

4C

ape

Ver

de P

ublic

Ent

erpr

ises

——

——

——

5C

entr

al A

fric

an R

epub

lic C

ivil

Serv

ice

88

——

——

6C

ongo

Civ

il Se

rvic

e—

——

——

—7

Eth

iopi

a M

ilita

ry54

2—

——

—54

28

Gha

na C

ivil

Serv

ice

2424

24—

——

9K

enya

Civ

il Se

rvic

e6

66

——

—10

Mal

awi C

ivil

Serv

ice

1010

10—

——

11M

auri

tani

a Pu

blic

Ent

erpr

ises

——

——

——

12N

amib

ia M

ilita

ry34

——

——

—13

Sene

gal C

ivil

Serv

ice

3030

——

——

14Si

erra

Leo

ne C

ivil

Serv

ice

11

1—

——

15U

gand

a C

ivil

Serv

ice

00

——

——

16U

gand

a M

ilita

ry17

1717

00

—

Asi

a17

Ban

glad

esh

Jute

Sec

tor

Publ

ic E

nter

pris

es18

18—

—0

—18

Cam

bodi

a C

ivil

Serv

ice

–126

–126

18—

144

—19

Chi

na S

heny

ang

Reg

ion

Ref

orm

33

30

0—

20In

dia

Publ

ic E

nter

pris

es83

8383

00

—21

Lao

PD

R C

ivil

Serv

ice

99

——

——

59

22Pa

kist

an P

ublic

Ent

erpr

ises

350

1818

——

332

23Pa

kist

an S

indh

Reg

ion

Ref

orm

——

——

——

24Sr

i Lan

ka C

ivil

Serv

ice

–157

–157

——

——

Eur

ope

25A

lban

ia P

ublic

Ent

erpr

ises

——

——

——

26H

unga

ry P

ublic

Ent

erpr

ises

298

298

——

——

27K

azak

hsta

n Pu

blic

Ent

erpr

ises

——

——

——

28M

aced

onia

Pub

lic E

nter

pris

es—

——

——

—29

Pola

nd P

ublic

Ent

erpr

ises

1,54

81,

148

1,14

8—

—40

030

Rus

sian

Fed

erat

ion

Coa

l Sec

tor

——

——

——

31T

urke

y Pu

blic

Ent

erpr

ises

——

——

——

Lat

in A

mer

ica

32A

rgen

tina

Pub

lic E

nter

pris

es23

723

723

7—

—0

33A

rgen

tina

Fed

eral

Adm

inis

trat

ion

1,06

31,

000

——

—63

34B

oliv

ia M

inin

g—Pu

blic

Cor

pora

tion

——

——

—0

35B

razi

l Civ

il Se

rvic

e0

0—

——

036

Chi

le C

ivil

Serv

ice

and

Para

stat

al O

rgan

izat

ions

——

——

——

37C

olom

bia

Tou

rism

and

Tra

nspo

rt M

inis

try

——

——

——

38E

cuad

or C

ivil

Serv

ice

——

——

——

39M

exic

o SO

CE

FI—

Min

istr

y of

Tra

de a

nd I

ndus

try

——

——

——

40Pe

ru C

ivil

Serv

ice

222

222

——

——

41Pe

ru S

UN

AT—

Tax

Col

lect

ing

Aut

hori

ty–1

5–1

51

016

—

— N

ot a

vaila

ble.

a. I

nclu

des

prog

ram

-spe

cifi

c ga

ins

such

as

a re

duct

ion

in a

rms

impo

rts

in E

thio

pia

and

rele

ase

of r

eal

esta

te i

n Pa

kist

an.

For

deta

ils,

see

indi

vidu

al c

ount

rysu

mm

arie

s (w

hich

are

ava

ilabl

e on

req

uest

fro

m t

he a

utho

rs).

Sour

ce: A

utho

rs’ c

alcu

lati

ons.

60

Tab

le A

-4.

Per

form

ance

Ind

icat

ors

for

Pub

lic S

ecto

r R

etre

nchm

ent

Pro

gram

s, b

y C

ount

ry a

nd P

rogr

am, 1

990s

Reg

ion,

cou

ntry

,B

reak

-eve

nN

et f

inan

cial

Pay

back

per

iod

Net

eco

nom

icN

umbe

ran

d ca

sepe

riod

(ye

ars)

gain

or

loss

(yea

rs)

gai

n or

loss

Afr

ica

1B

enin

Civ

il Se

rvic

e.—

–21

.——

2B

urki

na F

aso

Join

t R

ailw

ay L

ine

.——

.——

3C

amer

oon

Civ

il Se

rvic

e.—

—.—

—4

Cap

e V

erde

Pub

lic E

nter

pris

es.—

—.—

—5

Cen

tral

Afr

ican

Rep

ublic

Civ

il Se

rvic

e.—

—.—

—6

Con

go C

ivil

Serv

ice

.——

.——

7E

thio

pia

Mili

tary

0.40

342a

0.30

365a

8G

hana

Civ

il Se

rvic

e1.

82—

1.66

—9

Ken

ya C

ivil

Serv

ice

3.60

—3.

02—

10M

alaw

i Civ

il Se

rvic

e2.

00—

.——

11M

auri

tani

a Pu

blic

Ent

erpr

ises

.——

.——

12N

amib

ia M

ilita

ry1.

1048

b0.

4011

5b

13Se

nega

l Civ

il Se

rvic

e.—

–409

.——

14Si

erra

Leo

ne C

ivil

Serv

ice

1.87

—.—

—15

Uga

nda

Civ

il Se

rvic

e.—

–16

.——

16U

gand

a M

ilita

ry2.

70—

1.20

—

Asi

a17

Ban

glad

esh

Jute

Sec

tor

Publ

ic E

nter

pris

es3.

40—

.——

18C

ambo

dia

Civ

il Se

rvic

e.—

–143

6.—

—19

Chi

na S

heny

ang

Reg

ion

Ref

orm

.—29

0.00

9320

Indi

a Pu

blic

Ent

erpr

ises

.—–2

76.—

—21

Lao

PD

R C

ivil

Serv

ice

1.10

—.—

—22

Paki

stan

Pub

lic E

nter

pris

es1.

4450

1c.—

—

61

23Pa

kist

an S

indh

Reg

ion

Ref

orm

0.00

—.—

—24

Sri L

anka

Civ

il Se

rvic

e.—

–1,8

02.—

—

Eur

ope

25A

lban

ia P

ublic

Ent

erpr

ises

.——

.——

26H

unga

ry P

ublic

Ent

erpr

ises

.—–5

61.—

—27

Kaz

akhs

tan

Publ

ic E

nter

pris

es.—

—.—

—28

Mac

edon

ia P

ublic

Ent

erpr

ises

.——

.——

29Po

land

Pub

lic E

nter

pris

es.—

–6,1

21.—

—30

Rus

sian

Fed

erat

ion

Coa

l Sec

tor

.——

.——

31T

urke

y Pu

blic

Ent

erpr

ises

.——

.——

Lat

in A

mer

ica

32A

rgen

tina

Pub

lic E

nter

pris

es1.

56—

.——

33A

rgen

tina

Fed

eral

Adm

inis

trat

ion

0.41

—.—

—34

Bol

ivia

Min

ing.

—Pu

blic

Cor

pora

tion

10.0

0d—

10.0

0d—

35B

razi

l Civ

il Se

rvic

e.—

—.—

—36

Chi

le C

ivil

Serv

ice

and

Para

stat

al O

rgan

izat

ions

.——

.——

37C

olom

bia

Tou

rism

and

Tra

nspo

rt M

inis

try

.——

.——

38E

cuad

or C

ivil

Serv

ice

.——

.——

39M

exic

o SO

CE

FI.—

Min

istr

y of

Tra

de a

nd I

ndus

try

.——

.——

40Pe

ru C

ivil

Serv

ice

2.60

—.—

—41

Peru

SU

NA

T.—

Tax

Col

lect

ing

Aut

hori

ty.—

–47

0.00

2—

.— N

ot a

vaila

ble.

a. I

nclu

ding

sav

ings

in a

rms

impo

rts.

b. F

or t

hree

-yea

r pr

ogra

m d

urat

ion,

def

ense

exp

endi

ture

sav

ings

rev

erse

d th

erea

fter

.c.

Inc

ludi

ng p

riva

tiza

tion

pro

ceed

s.d.

Not

cal

cula

ted

(Wor

ld B

ank

1994

a).

Sour

ce: A

utho

rs’ c

alcu

lati

ons.

62

Tab

le A

-5.

Eco

nom

ic C

osts

and

Ben

efit

s of

Pub

lic S

ecto

r R

etre

nchm

ent

Pro

gram

s, b

y C

ount

ry a

nd P

rogr

am, 1

990s

(mill

ions

of

dolla

rs)

Eco

nom

ic c

osts

Eco

nom

ic b

enef

its

Reg

ion,

cou

ntry

,N

onfi

nanc

ial

Non

fina

ncia

l (p

rodu

ctio

n ri

se)

Num

ber

and

case

Tot

al(p

rodu

ctio

n lo

st)

Tot

alO

rgan

izat

ion

Wor

ker

Afr

ica

1B

enin

Civ

il Se

rvic

e21

——

——

2B

urki

na F

aso

Join

t R

ailw

ay L

ine

——

——

—3

Cam

eroo

n C

ivil

Serv

ice

7—

——

—4

Cap

e V

erde

Pub

lic E

nter

pris

es3

——

——

5C

entr

al A

fric

an R

epub

lic C

ivil

Serv

ice

——

8—

—6

Con

go C

ivil

Serv

ice

——

——

—7

Eth

iopi

a M

ilita

ry20

00

565

—23

8G

hana

Civ

il Se

rvic

e42

—26

—2

9K

enya

Civ

il Se

rvic

e20

—7

—1

10M

alaw

i Civ

il Se

rvic

e20

—10

——

11M

auri

tani

a Pu

blic

Ent

erpr

ises

9—

0—

—12

Nam

ibia

Mili

tary

38—

101

067

13Se

nega

l Civ

il Se

rvic

e80

—30

——

14Si

erra

Leo

ne C

ivil

Serv

ice

2—

1—

—15

Uga

nda

Civ

il Se

rvic

e16

—0

——

16U

gand

a M

ilita

ry42

—23

—5

Asi

a17

Ban

glad

esh

Jute

Sec

tor

Publ

ic E

nter

pris

es56

—18

——

18C

ambo

dia

Civ

il Se

rvic

e50

—–1

26—

—19

Chi

na S

heny

ang

Reg

ion

Ref

orm

00

1512

—20

Indi

a Pu

blic

Ent

erpr

ises

1,18

8—

83—

—

63

21L

ao P

DR C

ivil

Serv

ice

10—

9—

—22

Paki

stan

Pub

lic E

nter

pris

es25

—18

——

23Pa

kist

an S

indh

Reg

ion

Ref

orm

——

——

—24

Sri L

anka

Civ

il Se

rvic

e71

—–1

57—

—

Eur

ope

25A

lban

ia P

ublic

Ent

erpr

ises

6—

0—

—26

Hun

gary

Pub

lic E

nter

pris

es85

9—

298

——

27K

azak

hsta

n Pu

blic

Ent

erpr

ises

——

——

—28

Mac

edon

ia P

ublic

Ent

erpr

ises

50—

——

—29

Pola

nd P

ublic

Ent

erpr

ises

7,66

9—

1,14

8—

—30

Rus

sian

Fed

erat

ion

Coa

l Sec

tor

——

——

—31

Tur

key

Publ

ic E

nter

pris

es—

——

——

Lat

in A

mer

ica

32A

rgen

tina

Pub

lic E

nter

pris

es36

0—

237

Qua

ntit

ativ

e ri

ses

—33

Arg

enti

na F

eder

al A

dmin

istr

atio

n42

5—

1,00

0Q

uant

itat

ive

rise

s—

34B

oliv

ia M

inin

g—Pu

blic

Cor

pora

tion

74—

——

—35

Bra

zil C

ivil

Serv

ice

0—

00

036

Chi

le C

ivil

Serv

ice

and

Para

stat

alO

rgan

izat

ions

——

——

—37

Col

ombi

a T

ouri

sm a

nd T

rans

port

Min

istr

y—

——

——

38E

cuad

or C

ivil

Serv

ice

200

——

——

39M

exic

o SO

CE

FI—

Min

istr

y of

Tra

dean

d In

dust

ry—

——

——

40Pe

ru C

ivil

Serv

ice

530

—22

2—

—41

Peru

SU

NA

T—

Tax

Col

lect

ing

Aut

hori

ty2

093

394

7—

— N

ot a

vaila

ble.

Sour

ce: A

utho

rs’ c

alcu

lati

ons.


REFERENCES

The word “processed” describes informally reproduced works that may not be com-monly available through library systems. Included here are some references from theapproximately 200 pages of individual country case summaries that may not be cited inthe text.

Alderman, Harold, Roy S. Canagarajah, and Stephen Younger. 1996. “A Comparisonof Ghanaian Civil Servants’ Earnings before and after Retrenchment.” Journal ofAfrican Economies 4(2):259–88.

Basu, Kaushik, Gary Fields, and Shub Debgupta. 1996. “Retrenchment, Labor Laws,and Government Policy: An Analysis with Special Reference to India.” Paper pre-sented at the conference on Public Sector Retrenchment and Efficient CompensationSchemes, World Bank, Washington, D.C. Processed.

Bertola, Giuseppe. 1990. “Job Security, Employment, and Wages.” European EconomicReview 34(4, June):381–402.

Bertola, Giuseppe, and Richard Rogerson. 1997. “Institutions and Labor Reallocation.”European Economic Review 41(6, June):1147–71.

Blanchard, Olivier, and Peter Diamond. 1989. “The Beveridge Curve.” Brookings Pa-pers on Economic Activity 1:1–60.

———. 1990. “The Cyclical Behavior of Gross Flows of Workers in the United States.”Brookings Papers on Economic Activity 2:85–155.

Blanchard, Olivier, Simon Commander, and Fabrizio Coricelli. 1994. “Unemploymentand Restructuring in Eastern Europe.” In Simon Commander and Fabrizio Coricelli,eds., Unemployment, Restructuring, and the Labor Market in Eastern Europe andRussia. EDI Development Studies. Washington, D.C.: World Bank.

Commander, Simon, and Farbrizio Coricelli, eds. 1994. Unemployment, Restructuring,and the Labor Market in Eastern Europe and Russia. EDI Development Studies. Wash-ington, D.C.: World Bank.

Commander, Simon, Janos Kollo, Cecilia Ugaz, and Balacs Vilagi. 1994. “Hungary.” InSimon Commander and Fabrizio Coricelli, eds., Unemployment, Restructuring, andthe Labor Market in Eastern Europe and Russia. EDI Development Studies. Washing-ton, D.C.: World Bank.

Cooper, Russell, and John Haltiwanger. 1996. “Evidence on Macroeconomic Comple-mentarities.” Review of Economics and Statistics 78(1, February):78–93.

Davis, Steven J., John Haltiwanger, and Scott Schuh. 1996. Job Creation and Destruc-tion. Cambridge, Mass.: MIT Press.

Diwan, Ishac. 1993a. “Public Sector Retrenchment and Efficient Severance Pay Schemes.”World Bank, Policy Research Department, Washington, D.C. Processed.

———. 1993b. “Public Sector Retrenchment and Severance Pay: Nine Propositions.”World Bank, Policy Research Department, Washington, D.C. Processed.

Griliches, Zvi, ed. 1992. Output Measurement in the Service Sectors. NBER Studies inIncome and Wealth 56. Chicago: University of Chicago Press.

Hall, Robert E. 1995. “Lost Jobs.” Brookings Papers on Economic Activity 1:221–56.

International Monetary Fund. Various issues. International Finance Statistics. Washing-ton, D.C.


Lazear, Edward. 1990. “Job Security Provisions and Employment.” Quarterly Journalof Economics 105(3, August):699–726.

Levy, Anat, and Richard McLean. 1996. “Optimal and Suboptimal RetrenchmentSchemes: An Analytical Framework.” Paper presented at the conference on PublicSector Retrenchment and Efficient Compensation Schemes, World Bank, Washing-ton, D.C. Processed.

Lindauer, David, and Barbara Nunberg, eds. 1994. Rehabilitating Government: Payand Employment Reform in Africa. A World Bank Regional and Sectoral Study.Washington, D.C.: World Bank.

Mazumdar, Dipak. 1989. Microeconomic Issues of Labor Markets in Developing Coun-tries: Analysis and Policy Implications. EDI Seminar Paper 40. Washington, D.C.:World Bank.

Mills, Bradford, David E. Sahn, Edward E. Walden, and Stephen D. Younger. 1993.“Public Finance and Employment: An Analysis of Public Sector Retrenchment Pro-grams in Ghana and Guinea.” World Bank, Policy Research Department, Washing-ton, D.C. Processed.

OECD (Organisation for Economic Co-operation and Development). 1993. EmploymentOutlook (July). Paris.

———. 1996. Job Creation and Loss: Analysis, Policy and Data Development. Paris.PA Consulting Group. 1993. “Social Safety Net Adjustment Credit Technical Assis-

tance—Final Report.” Report submitted to the Government of India’s Industrial De-velopment Ministry and the World Bank. Auckland, New Zealand. Processed.

Rama, Martín. 1994. “Flexibility in Sri Lanka’s Labor Market.” Policy Research Work-ing Paper 1262. World Bank, Policy Research Department, Washington, D.C. Pro-cessed.

———. 1995. “Labor Market Cross-Country Database: Glossary.” World Bank, PolicyResearch Department, Washington, D.C. Processed.

———. 1997. “Efficient Public Sector Downsizing.” World Bank, Policy Research De-partment, Washington, D.C. Processed.

Robbins, Donald. 1996. “Public Sector Retrenchment: A Case Study of Argentina.”Paper presented at the conference on Public Sector Retrenchment and Efficient Com-pensation Schemes, World Bank, Washington, D.C. Processed.

Summers, Robert, and Alan Heston. 1991. “The Penn World Table (Mark 5): An Ex-panded Set of International Comparisons, 1950–1988.” Quarterly Journal of Eco-nomics 106(2, May):327–68.

Svejnar, Jan, and Katherine Terrell. 1991. “Reducing Labor Redundancy in State-OwnedEnterprises.” Policy Research Working Paper 792. World Bank, Policy Research De-partment, Washington, D.C. Processed.

Uganda Ministry of Public Service. 1994. Management of Change, Context, Vision,Objectives, Strategy, and Plan: Civil Service Reform. Kampala.

World Bank. 1994a. “Bolivia: Structural Reforms, Fiscal Impact, and Economic Growth.”Report 13067-B). Washington, D.C. Processed.

———. 1994b. “Peru Public Expenditure Review.” Report 13190-PE. Washington, D.C.Processed.

———. 1994c. “Shenyang Industrial Reform Project Appraisal Report.” Report 12415-CHA. Washington, D.C. Processed.


———. 1994d. World Development Report 1994: Infrastructure for Development. NewYork: Oxford University Press.

———. 1995a. “Argentina Public Sector Reform Project Completion Report.” Report14781. Washington, D.C. Processed.

———. 1995b. “From Swords to Ploughshares: The Uganda Demobilization and Rein-tegration of Ex-combatants in Uganda.” Washington, D.C. Processed.

———. 1995c. “Public Enterprise Adjustment Loan Audit Report.” Report 14809.Washington, D.C. Processed.

———. Various issues. World Tables. Baltimore, Md.: The Johns Hopkins UniversityPress.

Documents

Cross-Country Evidence on Public Sector Retrenchmentppiaf.org/sites/ppiaf.org/files/documents/toolkits/...John Haltiwanger and Manisha Singh are with the Department of Economics at